Bias Workshop, Reading, 8 - 11 November 2005 Satellite Instrument Calibration Issues: Geostationary platforms Ken Holmlund Meteorological Operations Division EUMETSAT [email protected]
Bias Workshop, Reading, 8 - 11 November 2005
Satellite Instrument Calibration Issues: Geostationary platforms
Ken HolmlundMeteorological Operations Division
Bias Workshop, Reading, 8 - 11 November 2005
Content• Introduction to SEVIRI (and MVIRI)• Blackbody calibration• Calibration issues and status• Clear Sky Radiance Product• Introduction to Atmospheric Motion Vectors• Reprocessing• Summary
Bias Workshop, Reading, 8 - 11 November 2005
MVIRI/SEVIRI Imaging Principle
– Earth imaging is obtained by bi-dimensional Earth scan combining:
• satellite spin at 100 rpm for East-West scan
• scan mirror motion for North-South scan
– SEVIRI scans 3 lines/revolution– MVIRI scans 1 line/revolution
Bias Workshop, Reading, 8 - 11 November 2005
calibration (about 2 sec)
Scan mirror rotation
Time (min)0 deg
+ 5.5 deg
- 5.5 deg
12 min 30 sec
15 min
retrace
Earth imaging
The image repeat cycle is split into the Earth imaging phase (1249 scan lines), the calibration phase (typically 4 lines) and the retrace phase (about 2’30” - profile driven by k-factors to inhibit nutation).
SEVIRI Imaging Principle
Bias Workshop, Reading, 8 - 11 November 2005
MVIRI vs. SEVIRI
• More frequent full disc imaging (15 min vs. 30 min)• Faster data transmission• More wavelengths (12 vs. 3)• Better sampling distance (normal:3 vs 4.5 km, HRV: 1 vs 2.25 km)• Note: Normal pixel size similar (4.8 vs. 5 km) (HRV: 1.4 vs. 2.25 km)
Bias Workshop, Reading, 8 - 11 November 2005
SEVIRI INSTRUMENTChannel Spectral Band in µm
λcen λmin λmax
Maximum Dynamic range
HRV Broadband (silicon response) 460 Wm-2sr-1µm-1
VIS0.6 0.635 0.56 0.71 533 Wm-2sr-1µm-1
VIS0.8 0.81 0.74 0.88 357 Wm-2sr-1µm-1
NIR1.6 1.64 1.50 1.78 75 Wm-2sr-1µm-1
IR3.9 3.90 3.48 4.36 335 KWV6.2 6.25 5.35 7.15 300 KWV7.3 7.35 6.85 7.85 300 KIR8.7 8.70 8.30 9.10 300 KIR9.7 9.66 9.38 9.94 310 KIR10.8 10.80 9.80 11.80 335 KIR12.0 12.00 11.00 13.00 335 KIR13.4 13.40 12.40 14.40 300 K
Bias Workshop, Reading, 8 - 11 November 2005
Calibration Status • Meteosat-8 (MSG-1)
– Black-body calibration
• Meteosat-7– Black-body calibration
• Meteosat-5 (Replaced by Meteosat-7 by end of 2006)– Cross-calibration with Meteosat-7 (May 2001)
• Meteosat-6– Image normalisation/correction with Meteosat-7 (May 2002)– Fixed calibration factor from IPS
Bias Workshop, Reading, 8 - 11 November 2005
Calibration issues
• The calibration unit is inserted in the optical path between the primary and secondary mirrors
• The effect from the from optics has to be modelled– 3 different models available, agree to within 0.1 - 0.5 K.
• Two point approach (cold and warm) • Non-linear effects are not considered• 2 - 3 hour oscillation in gain determines calibration
frequency• WV 6.2 micron channel gain setting
Bias Workshop, Reading, 8 - 11 November 2005
Applied corrections• Scan angle dependency (coating of the mirrors) for each channel
1.33
1.33
1.34
1.34
1.34
1.34
1.34
1.35
1.35
0 200 400 600 800 1000 1200 1400 1600SOL Counter
Red
uced
Gai
n
_1342 Retrace Trend Max Min 0.8%
Bias Workshop, Reading, 8 - 11 November 2005
The WV 6.2 micron channel gain setting
• Black body must be in the observable temperature range defined by the gain setting
• Current hot blackbody temperature is 320 K• WV 6.2 micron observations has therefore to cover
roughly a 120 K range (210 - 330 K)• Free atmosphere observations only to max 290 • Observation data 10 bit -> 8 bit, i.e. loss of sensitivity• Solution: change gain before and after BB-cal
– Currently done with MTP
• To be tested during MSG-2 commissioning
Bias Workshop, Reading, 8 - 11 November 2005
Calibration Validation
• Use of different blackbody calibration models– Agree to within 0.1 - 0.5 K
• Use of vicarious calibration– 7*7 pixel average at f/c grid or observation point– Minimum number of pixels = 40– Only clear sky over water for f/c
• Cross-calibration with other satellite data
Bias Workshop, Reading, 8 - 11 November 2005
GPRTM• Course resolution Line-by-line code
– Monochromatic calculation
• Input– Line parameters, surface emissivity– Model data, radiosonde
• Output– Up and downward radiance for IR channels of SEVIRI at different
levels in atmosphere for• SCE, AMV, ACT, CAL, TH
• Accuracy 1 - 2 % from comparison to reference calculations (LBL-RTM)
Bias Workshop, Reading, 8 - 11 November 2005
Vicarious calibration results with f/c12.0 micron: 0.222310 = .5% =.25K
13.4: 0.1576
Bias Workshop, Reading, 8 - 11 November 2005
WV Vicarious calibration results6.2 micron: Blue solid is forecast, Black is radiosonde showing problems
7.3 forecast only (lower panel)
5: 0.008318 6: 0.038622
Bias Workshop, Reading, 8 - 11 November 2005
Vicarious calibration resultsWith AIRS (Menzel)
0.28 K0.1 K1611 (13.4 µm)
0.11 K0.6 K1610 (12.1 µm)
0.09 K0.4 K169 (10.8 µm)
0.10 K-0.3 K168 (9.7 µm)
0.72 K-0.2 K167 (8.7 µm)
0.15 K-0.9 K166 (7.3 µm)
0.16 K-7.0 K165 (6.2 µm)
0.49 K-2.3 K164 (4.2 µm)
Stand. Dev.(from mean)
Mean ∆Tbb(MET-8 minus AIRS)
nBand
Bias Workshop, Reading, 8 - 11 November 2005
Clear Sky Radiance Product
• Pixel based cloud clearing• Calibrated mean radiances• 80 km resolution• Hourly dissemination
Bias Workshop, Reading, 8 - 11 November 2005
Clear Sky Radiance
CalibrationCloud clearingQuality control
Bias Workshop, Reading, 8 - 11 November 2005
Clear Sky Radiance ProductOpen issues
• All calibration related issues• Cloud contamination
– Gross errors– Diurnal variation (0.2 K)
• Quality Indicator
Bias Workshop, Reading, 8 - 11 November 2005
Image problems
• High inclination orbits– Currently all Meteosat First Generation (MFG)
• Eclipse effects– Mainly MFG
• Rotating lense– Meteosat-5
• Loose front optics– Meteosat-6– On-ground correction s/w– Currently using Meteosat-7 as reference
Bias Workshop, Reading, 8 - 11 November 2005
• Oscillation virtually gone• Small oscillations still on SH
- changed phase- rather related to model FG ?
NH
SH
High Inclination impact on CSR(“Wrong” sub-satellite point )
Correct viewing angle data provided as of 18 March 2003
Bias Workshop, Reading, 8 - 11 November 2005
Examples on eclipse effectsFor more examples and animations visit www.eumetsat.int
Bias Workshop, Reading, 8 - 11 November 2005
Examples on eclipse effectsFor more examples and animations visit www.eumetsat.int
Bias Workshop, Reading, 8 - 11 November 2005
Meteosat-5Rotating lense effect on observed radiance
1-2 June 1991 (Met-5 circles, Met-4 stars)
• Rotating lense (continuous)• Effect on Mean Earth count• Geometrical correction
applied for dissemination• Radiance variation 1-2%
during one image not corrected
Bias Workshop, Reading, 8 - 11 November 2005
Meteosat-6 Spurious gain changesweek 50 1996 (IR circles, WV diamonds)
• Spurious gain changes (up to 20%) during scanning • Effect on Mean Earth count (right)• Correction (left) applied for dissemination
Bias Workshop, Reading, 8 - 11 November 2005
ECMWF Radiance statistics (MET-8)100% clear (IR10.8 and 12.1 top, IR6.2 and 7.3 2 bottom)
Bias Workshop, Reading, 8 - 11 November 2005
ECMWF Radiance statisticsIR and WV, 100% clear (Met-7 top, Met-5 bottom)
Bias Workshop, Reading, 8 - 11 November 2005
MSG AMV Height Assignment
clrirRPop
irR
clrcoRPopcoRco
irclrirRirR
clrcoRcoR
cldirBfAwvLAwvcldwvB
cldirBclr
irL
cldirBclrwvLclrwvLclr
irLirwvopwvLop
wvLirwvcldirBclr
irL
opwvLclrwvL
irLirwvwvL
−
−=
−
−
=+
−
−−++
−
−=
)(
)(
)(
)1(
22222 ζ
τ
ζζζ
Bias Workshop, Reading, 8 - 11 November 2005
Automatic Quality Control
• Main goals: – Removal of gross errors– Indication of RMS error
• A set of consistency checks• Image correlation• Provision of a Quality Indicator• No real height assignment QI• Limited use for bias removal
Bias Workshop, Reading, 8 - 11 November 2005
MSG 10.8 AMV Collocations (NH, High levels)
0.00
0.10
0.20
0.30
0.40
0.5050100150200250
Horisontal Box Size (km)
NRM
S
0
8000
16000
24000
32000
40000
Num
ber o
f Col
loca
tions
NRMSNumber
Oper distance used 150 km
Collocation statisticsNRMS vs collocation box size
Bias Workshop, Reading, 8 - 11 November 2005
MSG 10.8 AMV Collocations (NH, High levels)
-1.50
-1.20
-0.90
-0.60
-0.30
0.0050100150200250
Box size (km)
Bia
s
0
8000
16000
24000
32000
40000
Num
bero
fCol
loca
tions
BiasNumber
Oper
Collocation statisticsBIAS vs collocation box size
Bias Workshop, Reading, 8 - 11 November 2005
NWP MonitoringMet Office bias: -0.8 m/s ECMWF bias -1.5 m/s
Bias Workshop, Reading, 8 - 11 November 2005
Reprocessing of Meteorological Products
• Using improved algorithms for better products• Support to e.g. ERA-40 and Interim-ERA• Re-Calibration of IR and WV Channels• Calibration of VIS channels
Bias Workshop, Reading, 8 - 11 November 2005
• 0 Degree Service• Meteosat-3: January till June 1989 Started• Meteosat-3: January till April 1990 Started• Meteosat-4: June 1989 – December 1989 Started• Meteosat-4: April 1990 till February 1994 completed• Meteosat-5: February 1994 till February 1997 completed• Meteosat-6: February 1997 till September 1997 completed• Meteosat-6: February 1997 till June 1998 completed, with the exception of October 1997• Meteosat-7: June 1998 till December 1999 completed• Meteosat-7: 2000 outstanding
• 63 Degree Service• Meteosat-5: June 1998 till December 2000 Started
• ADC Service• Meteosat-3: August 1991 till February 1993 Started
Bias Workshop, Reading, 8 - 11 November 2005
0 Degree ServiceMeteosat-3: 0 Degree Service: 75.0 % completedMeteosat-4: June 1989 till February 1994: 71 % completedMeteosat-5: February 1994 till February 1997: 87.7 % completedMeteosat-6: February 1997 till June 1998: 94.8 % completedMeteosat-7: June 1998 till December 2000: 34.7 % completed
63 Degree ServiceMeteosat-5: June 1998 till December 2000: 10.0 % completed
ADC ServiceMeteosat-3: August 1991 till February 1993 14 % completed
XADC Service August 1993 till May 1995 7 %
Bias Workshop, Reading, 8 - 11 November 2005
Summary• MSG calibration bias 0.5 - 1.5 K• MTP calibration shows significant biases (2 - 3 K)• Cloud contamination affects CSR products (.2 K diurnal
variation)
• AMV main source of error is height assignment• AMVs are not true point measurements• Use of forecast data enhances correlated errors?
• Rectification is generally within 1 pixel RMS, does not significantly impact NWP
Bias Workshop, Reading, 8 - 11 November 2005
Outlook• Calibration
– Adjustment of BB calibration model– Radiative transfer model
• Clear Sky Radiance Products– Improved cloud detection– Better image handling
• Atmospheric Motion Vectors– Improved Height assignment (Tuning, CO2, str, moisture corrections)– Height assignment quality indicators– Observation operator improvements